Language-agnostic text quality scorer that discriminates between clean UTF-8 text and mojibake, reversed text, wrong-codec decodings, and other corruption forms. Provides a standalone "languageyness" score suitable for re-OCR triggering and charset-decoding arbitration. Runtime classes (JunkDetector, ScriptDetector, feature extractors) and bundled model resources live here. Training and evaluation CLI tools live in the tools subpackage.